TerraInc

The 'TerraInc' dataset/benchmark is used to evaluate methods for federated domain generalization in an unsupervised setting, specifically focusing on the alignment of gradients to improve model generalization across different domains.

Papers using TerraInc (2)

Federated Domain Generalization with Label Smoothing and Balanced Decentralized Training2024Federated Unsupervised Domain Generalization using Global and Local Alignment of Gradients2024